Fig. 2: Reconstruction MSE vs number of qubits in the reservoir, with SL input coupling and different reservoir topologies.
From: State estimation with quantum extreme learning machines beyond the scrambling time
The evolution time in the upper panel (a) is t = 0.25 and in the lower panel (b) is t = 5. Each point represents the median of the MSEs computed on an ensemble of 500 random Hamiltonians, with sampling statistics of 106 for both training and testing in each case. Training and testing sets were each comprised of 50 random states. Errors bars represent first and third quartiles of the data. In addition to the standard topologies C, R, FC defined in Fig. 1, we also report the MSE obtained letting the whole input and reservoir system evolve unitarily with a Haar random unitary (RU).
